Jamaican music. Artist-led understanding.
Your words.
Your meaning.
Heard clearly.
Help AI hear Jamaican music the way you mean it.
InnaWords is a small, educational research project pairing songs with lyrics confirmed by the artists who made them. The goal: better transcription of Patois, Jamaican English and the meaning in between.
Contributions not open yet. Questions welcome.
Why this matters
The right words.
The real message.
A wrong word can change a whole line. Patois, fast delivery and cultural references can leave transcription tools guessing.
My hope is that, one day, the lyrics people read alongside music carry what the artist actually meant. This pilot starts by finding where existing tools get it wrong, using your own lyrics as the reference.
The aim is a 25-song pilot. Better captions and lyric timing are longer-term possibilities, not promised results.
Taking part
One song is a place to start.
- 01
Check the rights
Start with a song you control. Shared rights or an unclear contract? Let’s talk first.
- 02
Add your song + words
One audio file and the lyrics as you perform them. Keep your spelling, repeats and Patois.
- 03
Review, sign, then upload
Read the permissions before anything uploads. Keep a copy of your agreement.
WAV or FLAC preferred. A high quality MP3 works too. No stems or timestamps needed.
Your music stays yours
You keep ownership.
You choose the permissions.
- Limited permission to test transcription and review lyrics.
- No voice cloning, artist imitation or music generation.
- No commercial use or public sharing of your song or lyrics.
Model training needs separate permission and is currently disabled. Your original lyrics are preserved.
How your music is usedWhat you get back
A closer look at
how your song is heard.
For accepted songs that complete review, the intended return is:
- A reviewed lyric transcript.
- A short report on words the software missed.
- Experimental timed lyrics, where feasible.
There is no payment budget for this pilot. Review is manual, with no guaranteed delivery date.
Stefen Ewers is a Jamaican software engineer and Georgia Tech graduate student exploring how technology can better represent Jamaican language and culture.
I want our language to be understood on its own terms. This starts with listening to artists and keeping your words and meaning at the centre of the research.
Independent educational research, not sponsored by Georgia Tech or any music platform.
A few things you might ask
Before you share a song.
Can I take part if other people worked on the song?
Possibly. Co-writers, producers, featured artists, publishers and sample owners may also need to give permission. Contact Stefen first so the rights can be checked before you share any files.
Does using a distributor rule me out?
Not automatically. What matters is who controls the recording, music and lyrics, and whether your contracts allow the proposed research use.
Will this train an AI to sound like me?
No voice cloning, artist imitation or music generation is permitted. Testing transcription is the base research use. Training a speech-to-text model requires separate permission and remains disabled for now.
Can I change my mind?
You can request withdrawal. The agreement will explain how to stop future research use, request deletion and understand what limited records must be retained. Review those terms before contributing.
Will this put my lyrics on streaming platforms?
No. The hope is to contribute to better lyric tools over time. This pilot does not submit lyrics to platforms or guarantee placement, accuracy or delivery dates.
Let’s start a conversation
Have a song in mind?
We’re finishing the agreement and submission process. In the meantime, I’m happy to answer questions. Please don’t email audio or lyrics yet.